Employees are exploring AI. Is your organization ready for their ideas?

Executive summary: Employees can identify valuable opportunities for AI because they understand the work firsthand. Organizations need a clear path for evaluating those ideas, practical support during development, resources that make useful applications easy to find, and clear objectives for AI-enabled efficiency. Coordination between change management and AI delivery helps turn employee ideas into applications that improve specific processes and outcomes.

Employees are experimenting with AI and finding opportunities to improve their daily work. An employee who spends hours on a repetitive process may see an opportunity that would never emerge from a top-down search for AI applications.

Recognizing an opportunity, however, is only the beginning. Employees need to know where to take an idea, how to develop it, and where to get help when they encounter a problem. Organizations also need a consistent way to determine what problem an application would solve, whether the idea merits investment, and how they will know whether it worked.

Change management and AI delivery address different parts of the challenge. Change management helps employees participate, learn, and adopt new applications. AI delivery provides the processes and technical support needed to evaluate, develop, and manage those applications. Coordinating the two gives employee ideas a practical path from initial exploration through everyday use.

Our work with a large utility shows what the approach can look like in practice.

Giving employees a defined path

Microsoft Copilot was already available at the utility, and teams were experimenting with it. Employees had ideas for new agents, but requests arrived through informal channels. Adoption remained concentrated among enthusiasts, and the organization lacked a consistent process for evaluating proposals.

Logic20/20 continues to support the utility's AI program through a structured intake process and coordinated change management efforts:

• Through a structured request process, teams evaluate proposals based on the problem each idea addresses, the outcome it is expected to improve, feasibility, data readiness, and governance requirements, then direct viable ideas to the appropriate builders.
• A shared backlog gives leaders visibility into proposed work and helps them decide where development capacity can have the greatest impact.

At the same time, workshops, build-a-thons, and AI champions within business units encourage more employees to participate. Published examples show employees how colleagues are applying AI and give them ideas for their own work.

Establishing the intake process while encouraging employee participation addresses an important dependency. Encouraging employees to experiment creates demand for development support. An intake process works only when employees know about it and understand how to participate.

As participation grows, change management and delivery need to remain connected:

• Delivery teams use employee demand to refine development resources and support.
• Change teams use information from delivery teams to give employees accurate guidance about available applications, development options, and requirements.

Support learning when employees need it

Formal training can introduce employees to AI, but much of the learning happens once they begin applying it to their own work.

At the utility, weekly office hours give employees a place to troubleshoot agents with experienced developers. Employees arrive with specific problems from their own builds and receive guidance they can apply immediately. Logic20/20 developers provide direct assistance when employees encounter more complex issues.

The experience shows why development support matters alongside formal training. Employees often need help at the moment they are applying AI to actual work, not just when a training session happens to be scheduled. When employees encounter a problem while building an application, the instruction has immediate context. Solving the problem helps advance the application while building knowledge the employee can use again.

Organizations can reinforce learning through complementary forms of support:

• Direct development assistance. Office hours and experienced developers help employees work through specific challenges while building their applications.
• Reusable components and examples. Instead of beginning every build from scratch, employees can start with established patterns and focus their effort on the needs of their particular workflow.
• Local trainers and role-based learning resources. Trainers within a business unit understand its operating context and can address questions in familiar terms. Reference materials give employees a resource they can revisit when a need arises weeks or months after formal training.

The goal is not simply to teach employees about AI. Development support helps employees apply what they learn to problems they already understand.

Make useful tooling easier to find

As the number of available agents grows, organizations face a different adoption challenge: helping employees identify which tools are relevant to their work.

A large catalog does not necessarily make AI more accessible. Employees may see dozens of available agents without knowing which one fits a particular task or whether trying one will be worth their time.

Organizing resources around roles and workflows makes the choice easier. Employees need enough context to answer three questions before they begin:

• What work does the agent support? Describe its purpose using language employees recognize from their responsibilities.
• What does using it look like? Previews and short videos can show how the agent works and what employees can expect.
• How have colleagues used it? Examples and internal success stories show where an agent has already been applied and what employees accomplished with it.

Learning resources benefit from the same approach. Organizing guidance by role and task helps employees find instruction relevant to the work in front of them rather than searching through a general library.

Discovery also requires ongoing attention. Every newly published agent changes the set of options available to employees. Communications, learning resources, and catalogs need to reflect what is actually available so employees can make informed choices.

Define the purpose of AI-enabled efficiency

Employee participation also depends on how the organization talks about productivity.

Employees experimenting with AI may find ways to complete work in less time. They have a legitimate interest in what the organization intends to do with the capacity AI creates. Without clear direction, employees may draw their own conclusions, including concerns about job security.

Leaders can reduce ambiguity by defining the business objective behind efficiency efforts. An organization may intend to support growth, increase capacity for priority work, improve service, or advance another established objective. A stated objective gives employees context for why the organization is asking them to identify more efficient ways of working and what the resulting capacity is intended to accomplish.

Organizations may not have every workforce question resolved when AI programs begin. In those situations, acknowledging uncertainty is more credible than offering assurances leadership has not established. Employees still benefit from knowing what the organization is trying to accomplish and how leaders will evaluate progress.

Measure outcomes, not just time saved

Time saved is a useful starting point, but it does not tell leaders what changed as a result. Pair hours unlocked with the measures the business already uses to evaluate performance. Depending on the process, that could mean cycle time, cost, reliability, quality, risk, customer experience, or throughput. The question shifts from “How much time did AI save?” to “What did the organization accomplish with the capacity it created?”

Connect participation with operational improvement

Employee interest in AI creates an opportunity, but interest alone does not reduce workload, improve a process, or advance an organizational objective. Organizations need the capacity to act on what employees know about their work and connect promising ideas with the outcomes leaders are accountable for improving.

Supporting employee ideas takes more than training or an intake process in isolation. Employees need somewhere to take an idea, practical help developing it, and an easy way to find and reuse what already exists. Leaders need criteria for deciding which ideas warrant development and which outcomes will show whether they worked.

That is where change management and AI delivery intersect. Change efforts bring employees into the process and help useful applications become part of everyday work; delivery teams give promising ideas a path through evaluation, development, governance, and support. Information from each should continually shape the other.

Employees can then contribute ideas knowing what will happen to them, while leaders can direct development toward defined outcomes: less repetitive work, shorter cycle times, greater reliability, or more capacity for priority work. Employee experimentation becomes a source of operational improvement rather than a collection of disconnected AI experiments.

Learn more about scaling AI adoption